The Mobius AIRO mobile CT for image-guided proton therapy: Characterization & commissioning

Open-source deep-learning models for segmentation of normal structures for prostatic and gynecological high-dose-rate brachytherapy: Comparison of architectures

BACKGROUND

The use of deep learning-based auto-contouring algorithms in various treatment planning services is increasingly common. There is a notable deficit of commercially or publicly available models trained on large or diverse datasets containing high-dose-rate (HDR) brachytherapy treatment scans, leading to poor performance on images that include HDR implants.

PURPOSE

To implement and evaluate automatic organs-at-risk (OARs) segmentation models for use in prostatic-and-gynecological computed tomography (CT)-guided high-dose-rate brachytherapy treatment planning.

METHODS AND MATERIALS

1316 computed tomography (CT) scans and corresponding segmentation files from 1105 prostatic-or-gynecological HDR patients treated at our institution from 2017 to 2024 were used for model training. Data sources comprised six CT scanners including a mobile CT unit with previously reported susceptibility to image streaking artifacts. Two UNet-derived model architectures, UNet++ and nnU-Net, were investigated for bladder and rectum model training. The models were tested on 100 CT scans and clinically-used segmentation files from 62 prostatic-or-gynecological HDR brachytherapy patients, disjoint from the training set, collected in 2024. Performance was evaluated using the Dice-Similarity-Coefficient (DSC) between model predicted contours and clinically-used contours on slices in common with the Clinical-Target-Volume (CTV). Additionally, a blinded evaluation of ten random test cases was conducted by three experienced planners.

RESULTS

Median (interquartile range) 3D DSC on CTV-containing slices were 0.95 (0.04) and 0.87 (0.09) for the UNet++ bladder and rectum models, respectively, and 0.96 (0.03) and 0.88 (0.10) for the nnU-Net. The rank-sum test did not reveal statistically significant differences in these DSC (p = 0.15 and 0.27, respectively). The blinded evaluation scored trained models higher than clinically-used contours.

CONCLUSION

Both UNet-derived architectures perform similarly on the bladder and rectum and are adequately accurate to reduce contouring time in a review-and-edit context during HDR brachytherapy planning. The UNet++ models were chosen for implementation at our institution due to lower computing hardware requirements and are in routine clinical use.

A review of the clinical introduction of 4D particle therapy research concepts

Background and purpose

Many 4D particle therapy research concepts have been recently translated into clinics, however, remaining substantial differences depend on the indication and institute-related aspects. This work aims to summarise current state-of-the-art 4D particle therapy technology and outline a roadmap for future research and developments.

Material and methods

This review focused on the clinical implementation of 4D approaches for imaging, treatment planning, delivery and evaluation based on the 2021 and 2022 4D Treatment Workshops for Particle Therapy as well as a review of the most recent surveys, guidelines and scientific papers dedicated to this topic.

Results

Available technological capabilities for motion surveillance and compensation determined the course of each 4D particle treatment. 4D motion management, delivery techniques and strategies including imaging were diverse and depended on many factors. These included aspects of motion amplitude, tumour location, as well as accelerator technology driving the necessity of centre-specific dosimetric validation. Novel methodologies for X-ray based image processing and MRI for real-time tumour tracking and motion management were shown to have a large potential for online and offline adaptation schemes compensating for potential anatomical changes over the treatment course. The latest research developments were dominated by particle imaging, artificial intelligence methods and FLASH adding another level of complexity but also opportunities in the context of 4D treatments.

Conclusion

This review showed that the rapid technological advances in radiation oncology together with the available intrafractional motion management and adaptive strategies paved the way towards clinical implementation.

Technical note: Characterization of novel iterative reconstructed cone beam CT images for dose tracking and adaptive radiotherapy on L-shape linacs

BACKGROUND

Cone-beam computed tomography (CBCT) allows for patient setup and positioning, and potentially dose verification or adaptive replanning prior to each treatment delivery. Poor CBCT image quality due to scatter artifacts and patient motion has been a major limiting factor. A new image reconstruction algorithm was recently clinically implemented for improving image quality through iterative reconstruction (iCBCT).

PURPOSE

This study aims to characterize iCBCT image quality, establish image value (HU)-to-relative electron density (RED) calibration curves for dose calculation, and assess the dosimetric accuracy for different anatomical sites.

MATERIAL AND METHODS

Both conventional CBCT and iCBCT scans were acquired from a Varian TrueBeam On-Board Imager system. A Catphan 604 phantom was scanned to compare image quality between the traditional Feldkamp-Davis-Kress (FDK) and novel iterative reconstruction techniques. Computerized Imaging Reference Systems (CIRS) electron density phantom was used to construct site-specific HU-RED curves corresponding to various scan settings. The CIRS Dynamic Thorax phantom, Rando pelvis phantom, and BrainLab head phantom were used for assessing dosimetric accuracy calculated on iCBCT images, compared to that on traditional FDK-based CBCT images. All phantoms were scanned on a computed tomography (CT) to obtain baseline HU values for comparison.

RESULTS

Test results obtained from Catphan showed statistically significant improvement with iCBCT, compared to FDK CBCT. Average HU differences from the baseline CT values were improved to within ±30 HU for iCBCT, compared to FDK CBCT for phantom studies. Dose calculated on iCBCT for both phantoms and patient cases directly using baseline HU-RED calibration from CT showed 0.5%-2.0% accuracy from the baseline dose calculated on CT, which is comparable to doses calculated using site-specific HU-RED calibration curves.

CONCLUSION

iCBCT provides improved image quality, improved HU accuracy compared to CT baseline, and has potential to provide online dose verification as part of the adaptive radiotherapy workflow directly using the baseline HU-RED calibration curve from CT.

Effect of subject motion and gantry rotation speed on image quality and dose delivery in CT-guided radiotherapy

PURPOSE

To investigate the effects of subject motion and gantry rotation speed on computed tomography (CT) image quality over a range of image acquisition speeds for fan-beam (FB) and cone-beam (CB) CT scanners, and quantify the geometric and dosimetric errors introduced by FB and CB sampling in the context of adaptive radiotherapy.

METHODS

Images of motion phantoms were acquired using four CT scanners with gantry rotation speeds of 0.5 s/rotation (denoted FB-0.5), 1.9 s/rotation (FB-1.9), 16.6 s/rotation (CB-16.6), and 60.0 s/rotation (CB-60.0). A phantom presenting various tissue densities undergoing motion with 4-s period and ranging in amplitude from ±0.5 to ±10.0 mm was used to characterize motion artifacts (streaks), motion blur (edge-spread function, ESF), and geometric inaccuracy (excursion of insert centroids and distortion of known shape). An anthropomorphic abdomen phantom undergoing ±2.5-mm motion with 4-s period was used to simulate an adaptive radiotherapy workflow, and relative geometric and dosimetric errors were compared between scanners.

RESULTS

At ±2.5-mm motion, phantom measurements demonstrated mean ± SD ESF widths of 0.6 ± 0.0, 1.3 ± 0.4, 2.0 ± 1.1, and 2.9 ± 2.0 mm and geometric inaccuracy (excursion) of 2.7 ± 0.4, 4.1 ± 1.2, 2.6 ± 0.7, and 2.0 ± 0.5 mm for the FB-0.5, FB-1.9, CB-16.6, and CB-60.0 scanners, respectively. The results demonstrated nonmonotonic trends with scanner speed for FB and CB geometries. Geometric and dosimetric errors in adaptive radiotherapy plans were largest for the slowest (CB-60.0) scanner and similar for the three faster systems (CB-16.6, FB-1.9, and FB-0.5).

CONCLUSIONS

Clinically standard CB-60.0 demonstrates strong image quality degradation in the presence of subject motion, which is mitigated through faster CBCT or FBCT. Although motion blur is minimized for FB-0.5 and FB-1.9, such systems suffer from increased geometric distortion compared to CB-16.6. Each system reflects tradeoffs in image artifacts and geometric inaccuracies that affect treatment delivery/dosimetric error and should be considered in the design of next-generation CT-guided radiotherapy systems.

Operational Performance of a Compact Proton Therapy System: A 5-Year Experience

Purpose

We present an analysis of various operational metrics for a novel compact proton therapy system, including clinical case mix, subsystems utilization, and quality assurance trends in beam delivery parameters over a period of 5 years.

Materials and Methods

Patient-specific data from a total of 850 patients (25,567 fractions) have been collected and analyzed. The patient mix include a variety of simple, intermediate, and complex cases. Beam-specific delivery parameters for a total of 3585 beams were analyzed. In-room imaging system usage for off-line adaptive purpose is reported. We also report key machine performances metrics based on routine quality assurance in addition to uptime.

Results

Our analysis shows that system subcomponents including gantry and patient positioning system have maintained a tight mechanical tolerance over the 5-year period. Various beam parameters were all within acceptable tolerances with no clear trends. Utilization frequency histograms of gantry and patient positioning system show that only a small fraction of all available angles was used for patient deliveries with cardinal angels as the most usable. Similarly, beam-specific metrics, such as range, modulation, and air gaps, were clustered unevenly over the available range indicating that this compact system was more than capable to treat the complex variety of tumors of our patient mix.

Conclusion

Our data show that this compact system is versatile, robust, and capable of delivering complex treatments like a large full-gantry system. Utilization data show that a fraction of all subcomponents range of angular motion has been used. Compilation of beam-specific metrics, such as range and modulation, show uneven distributions with specific clustering over the entire usable range. Our findings could be used to further optimize the performance and cost-effectiveness of future compact proton systems.

Technological Advances in Charged-Particle Therapy

Charted-particle therapy (CPT) benefits cancer patients by localizing doses in the tumor volume while minimizing the doses delivered to normal tissue through its unique physical and biological characteristics. The world's first CPT applied on humans was proton beam therapy (PBT), which was performed in the mid-1950s. Among heavy ions, carbon ions showed the most favorable biological characteristics for the treatment of cancer patients. Carbon ions show coincidence between the Bragg peak and maximum value of relative biological effectiveness. In addition, they show low oxygen enhancement ratios. Therefore, carbon-ion radiotherapy (CIRT) has become mainstream in the treatment of cancer patients using heavy ions. CIRT was first performed in 1977 at the Lawrence Berkeley Laboratory. The CPT technology has advanced in the intervening decades, enabling the use of rotating gantry, beam delivery with fast pencil-beam scanning, image-guided particle therapy, and intensity-modulated particle therapy. As a result, as of 2019, a total of 222,425 and 34,138 patients with cancer had been treated globally with PBT and CIRT, respectively. For more effective and efficient CPT, many groups are currently conducting further studies worldwide. This review summarizes recent technological advances that facilitate clinical use of CPT.

Effective dose and image quality for intraoperative imaging with a cone-beam CT and a mobile multi-slice CT in spinal surgery: A phantom study

PURPOSE

To compare the effective dose (ED) and image quality (IQ) of O-arm cone-beam CT (Medtronic, Minneapolis, MN, USA) and Airo multi-slice CT (Brainlab AG, Munich, Germany) for intraoperative-CT (i-CT) in spinal surgery.

METHODS

The manufacturer-defined protocols available in the O-arm and Airo systems for three-dimensional lumbar spine imaging were compared. Organ dose was measured both with thermo-luminescent dosimeters and GafChromic films in the Alderson RadiationTherapy anthropomorphic phantom. A subjective analysis was performed by neurosurgeons to compare the clinical IQ of the anthropomorphic phantom images acquired with the different i-CT systems and imaging protocols. Image uniformity, noise, contrast-to-noise-ratio (CNR), and spatial resolution were additionally assessed with the Catphan 504 phantom.

RESULTS

O-arm i-CT caused 56% larger ED than Airo due to the high definition (HD) imaging protocol. The noise was larger for O-arm images leading to a lower CNR than that measured for Airo. Moreover, scattering and beam hardening effects were observed in the O-arm images. Better spatial resolution was measured for the O-arm system (9 lp/cm) than for Airo (4 lp/cm). For all the investigated protocols, O-arm was found to be better for identifying anatomical features important for accurate pedicle screw positioning.

CONCLUSIONS

According to phantom measurements, the HD protocol of O-arm offered better clinical IQ than Airo but larger ED. The larger noise of O-arm images did not compromise the clinical IQ while the superior spatial resolution of this system allowed a better visibility of anatomical features important for pedicle screw positioning in the lumbar region.

A novel approach to Verify air gap and SSD for proton radiotherapy using surface imaging

PURPOSE

To develop a novel approach to accurately verify patient set up in proton radiotherapy, especially for the verification of the nozzle - body surface air gap and source-to-skin distance (SSD), the consistency and accuracy of which is extremely important in proton treatment.

METHODS

Patient body surfaces can be captured and monitored with the optical surface imaging system during radiation treatment for improved intrafraction accuracy. An in-house software package was developed to reconstruct the patient body surface in the treatment position from the optical surface imaging reference capture and to calculate the corresponding nozzle - body surface air gap and SSD. To validate this method, a mannequin was scanned on a CT simulator and proton plans were generated for a Mevion S250 Proton machine with 20 gantry/couch angle combinations, as well as two different snout sizes, in the Varian Eclipse Treatment Planning Systems (TPS). The surface generated in the TPS from the CT scan was imported into the optical imaging system as an RT Structure for the purpose of validating and establishing a benchmark for ground truth comparison. The optical imaging surface reference capture was acquired at the treatment setup position after orthogonal kV imaging to confirm the positioning. The air gaps and SSDs calculated with the developed method from the surface captured at the treatment setup position (VRT surface) and the CT based surface imported from the TPS were compared to those calculated in TPS. The same approach was also applied to 14 clinical treatment fields for 10 patients to further validate the methodology.

RESULTS

The air gaps and SSDs calculated from our program agreed well with the corresponding values derived from the TPS. For the phantom results, using the CT surface, the absolute differences in the air gap were 0.45 mm ± 0.33 mm for the small snout, and 0.51 mm ± 0.49 mm for the large snout, and the absolute differences in SSD were 0.68 mm ± 0.42 mm regardless of snout size. Using the VRT surface, the absolute differences in air gap were 1.17 mm ± 1.17 mm and 2.1 mm ± 3.09 mm for the small and large snouts, respectively, and the absolute differences in SSD were 0.81 mm ± 0.45 mm. Similarly, for patient data, using the CT surface, the absolute differences in air gap were 0.42 mm ± 0.49 mm, and the absolute differences in SSD were 1.92 mm ± 1.4 mm. Using the VRT surface, the absolute differences in the air gap were 2.35 mm ± 2.3 mm, and the absolute differences in SSD were 2.7 mm ± 2.17 mm.

CONCLUSION

These results showed the feasibility and robustness of using an optical surface imaging approach to conveniently determine the air gap and SSD in proton treatment, providing an accurate and efficient way to confirm the target depth at treatment.

Technical assessment of a mobile CT scanner for image-guided brachytherapy

PURPOSE

The imaging performance and dose of a mobile CT scanner (Brainlab Airo®, Munich, Germany) is evaluated, with particular consideration to assessment of technique protocols for image-guided brachytherapy.

METHOD

Dose measurements were performed using a 100-mm-length pencil chamber at the center and periphery of 16- and 32-cm-diameter CTDI phantoms. Hounsfield unit (HU) accuracy and linearity were assessed using materials of specified electron density (Gammex RMI, Madison, WI), and image uniformity, noise, and noise-power spectrum (NPS) were evaluated in a 20-cm-diameter water phantom as well as an American College of Radiology (ACR) CT accreditation phantom (Model 464, Sun Nuclear, Melbourne, FL). Spatial resolution (modulation transfer function, MTF) was assessed with an edge-spread phantom and visually assessed with respect to line-pair patterns in the ACR phantom and in structures of interest in anthropomorphic phantoms. Images were also obtained on a diagnostic CT scanner (Big Bore CT simulator, Philips, Amsterdam, Netherlands) for qualitative and quantitative comparison. The manufacturer's metal artifact reduction (MAR) algorithm was assessed in an anthropomorphic body phantom containing surgical instrumentation. Performance in application to brachytherapy was assessed with a set of anthropomorphic brachytherapy phantoms - for example, a vaginal cylinder and interstitial ring and tandem.

RESULT

Nominal dose for helical and axial modes, respectively, was 56.4 and 78.9 mGy for the head protocol and 17.8 and 24.9 mGy for the body protocol. A high degree of HU accuracy and linearity was observed for both axial and helical scan modes. Image nonuniformity (e.g., cupping artifact) in the transverse (x,y) plane was less than 5 HU, but stitching artifacts (~5 HU) in the longitudinal (z) direction were observed in axial scan mode. Helical and axial modes demonstrated comparable spatial resolution of ~5 lp/cm, with the MTF reduced to 10% at ~0.38 mm-1 . Contrast-to-noise ratio was suitable to soft-tissue visualization (e.g., fat and muscle), but windmill artifacts were observed in helical mode in relation to high-frequency bone and metal. The MAR algorithm provided modest improvement to image quality. Overall, image quality appeared suitable to relevant clinical tasks in intracavitary and interstitial (e.g., gynecological) brachytherapy, including visualization of soft-tissue structures in proximity to the applicators.

CONCLUSION

The technical assessment highlighted key characteristics of dose and imaging performance pertinent to incorporation of the mobile CT scanner in clinical procedures, helping to inform clinical deployment and technique protocol selection in brachytherapy. For this and other possible applications, the work helps to identify protocols that could reduce radiation dose and/or improve image quality. The work also identified areas for future improvement, including reduction of stitching, windmill, and metal artifacts.

The Dosimetric Impact of Interfractional Organ-at-Risk Movement During Liver Stereotactic Body Radiation Therapy

PURPOSE

Stereotactic body radiation therapy (SBRT) is an effective therapy for treating liver malignancies. However, little is known about interfractional dose variations to adjacent organs at risk (OARs). We examine the effects of interfractional organ movement and setup variation on dose delivered to OARs in patients receiving liver SBRT.

METHODS AND MATERIALS

Thirty patients treated with liver SBRT were analyzed. Daily image guidance with diagnostic quality computed tomography-on-rails imaging was performed before each fraction. In phase 1, these daily images were used to delineate all OARs including the liver, heart, right kidney, esophagus, stomach, duodenum, and large bowel in 10 patients. In phase 2, only OARS in close proximity to the target were contoured in 20 additional patients. Dose distribution on each daily computed tomography was generated, and daily doses to each OAR were recorded and compared with clinical thresholds to determine whether a daily dose excess (DDE) occurred.

RESULTS

In phase 1, significant interfractional dose differences between planned and delivered dose to OARs were observed, but differences were rarely clinically significant, with just 1 DDE. In phase 2, multiple DDEs were recorded for OARs close to the target, mainly involving the stomach, heart, and esophagus. Tumors in the hilum and liver segments I, IV, and VIII were the most common locations for DDEs. On root cause analysis, 3 etiologies of DDE emerged: craniocaudal shift (69.2%), anatomic changes (28.2%), and anteroposterior shifts (2.6%).

CONCLUSIONS

OARs close to liver lesions may receive higher doses than expected during SBRT owing to interfractional variations in OARs relative to the target. These differences in planned versus expected dose can lead to toxicity. Efforts to better evaluate OARs with daily image guidance may help reduce risks. Application of adaptive replanning and improved and real-time image guidance could mitigate risks of toxicity, and further study into their applications is warranted.

Current state and future applications of radiological image guidance for particle therapy

In this review paper, we first give a short overview of radiological image guidance in photon radiotherapy, placing emphasis on the fact that linac based radiotherapy has outpaced particle therapy in the adoption of volumetric image guidance. While cone beam computed tomography (CBCT) has been an established technique in linac treatment rooms for almost two decades, the widespread adoption of volumetric image guidance in particle therapy, whether by means of CBCT or in-room CT imaging, is recent. This lag may be attributable to the bespoke nature and lower number of particle therapy installations, as well as the differences in geometry between those installations and linac treatment rooms. In addition, for particle therapy the so called shift invariance of the dose distribution rarely applies. An overview of the different volumetric image guidance solutions found at modern particle therapy facilities is provided, covering gantry, nozzle, C-arm, and couch-mounted CBCT as well different in-room CT configurations. A summary of the use of in-room volumetric imaging data beyond anatomy-based positioning is also presented as well as the necessary corrections to CBCT images for accurate water equivalent thickness calculation. Finally, the use of non-ionizing imaging modalities is discussed.

Commissioning an in-room mobile CT for adaptive proton therapy with a compact proton system

Purpose

To describe the commissioning of AIRO mobile CT system (AIRO) for adaptive proton therapy on a compact double scattering proton therapy system.

Methods

A Gammex phantom was scanned with varying plug patterns, table heights, and mAs on a CT simulator (CT Sim) and on the AIRO. AIRO‐specific CT‐stopping power ratio (SPR) curves were created with a commonly used stoichiometric method using the Gammex phantom. A RANDO anthropomorphic thorax, pelvis, and head phantom, and a CIRS thorax and head phantom were scanned on the CT Sim and AIRO. Clinically realistic treatment plans and nonclinical plans were generated on the CT Sim images and subsequently copied onto the AIRO CT scans for dose recalculation and comparison for various AIRO SPR curves. Gamma analysis was used to evaluate dosimetric deviation between both plans.

Results

AIRO CT values skewed toward solid water when plugs were scanned surrounded by other plugs in phantom. Low‐density materials demonstrated largest differences. Dose calculated on AIRO CT scans with stoichiometric‐based SPR curves produced over‐ranged proton beams when large volumes of low‐density material were in the path of the beam. To create equivalent dose distributions on both data sets, the AIRO SPR curve's low‐density data points were iteratively adjusted to yield better proton beam range agreement based on isodose lines. Comparison of the stoichiometric‐based AIRO SPR curve and the “dose‐adjusted” SPR curve showed slight improvement on gamma analysis between the treatment plan and the AIRO plan for single‐field plans at the 1%, 1 mm level, but did not affect clinical plans indicating that HU number differences between the CT Sim and AIRO did not affect dose calculations for robust clinical beam arrangements.

Conclusion

Based on this study, we believe the AIRO can be used offline for adaptive proton therapy on a compact double scattering proton therapy system.

A pencil beam algorithm for magnetic resonance image-guided proton therapy

Purpose

The feasibility of magnetic resonance image (MRI)‐based proton therapy is based, among several other factors, on the implementation of appropriate extensions on current dose calculation methods. This work aims to develop a pencil beam algorithm (PBA) for dose calculation of proton beams within magnetic field regions of up to 3 T.

Methods

Monte Carlo (MC) simulations using the GATE 7.1/GEANT4.9.4p02 toolkit were performed to generate calibration and benchmarking data for the PBA. Dose distributions from proton beams in the clinical required energy range 60–250 MeV impinging on a 400 × 400 × 400 mm³ water phantom and transverse magnetic fields ranging from 0 to 3 T were considered. Energy depositions in homogeneous and heterogeneous phantoms filled with water, adipose, bone, and air were evaluated for proton energies of 80, 150, and 240 MeV, combining a trajectory calculation method and look‐up tables (LUT). A novel parametrization model, using independent tailed Gauss fitting functions, was employed to describe the nonsymmetric shape of lateral beam profiles. Integrated depth‐dose curves (IDD), lateral dose profiles, and two‐dimensional dose distributions calculated with the PBA were compared with results from MC simulations to assess the performance of the algorithm. A gamma index criterion of 2%/2 mm was used for analysis.

Results

A close to perfect agreement was observed for PB‐based dose calculations in water in magnetic fields of 0.5, 1.5, and 3 T. IDD functions showed differences between the PBA and MC of less than 0.1% before the Bragg peak, and deviations of 2–8% in the distal energy falloff region. Gamma index pass rates higher than 99% and mean values lower than 0.1 were encountered for all analyzed configurations. For homogeneous phantoms, only the full bone configuration offered deviations in the Bragg peak position of up to 1.7% and overestimations of the lateral beam spot width for high‐energy protons and magnetic field intensities. An excellent agreement between PBA and MC dose calculation was also achieved using slab‐like and lateral heterogeneous phantoms, with gamma index passing rates above 98% and mean values between 0.1 and 0.2. As expected, agreement reduced for high‐energy protons and high‐intensity magnetic fields, although results remained good enough to be considered for future implementation in clinical practice.

Conclusions

The proposed pencil beam algorithm for protons can accurately account for dose distortion effects induced by external magnetic fields. The application of an analytical model for dose estimation and corrections reduces the calculation times considerably, making the presented PBA a suitable candidate for integration in a treatment planning system.